On Mon, 29 Aug 2011 18:14:49 -0700, Jesse Molina wrote:
Thanks Jesse, your post was very informational, if you should feel in
the mood to publish that forum post, i'd be very interested in reading
it.
I had netgraph visible while I was playing, and didn't really notice
anything spectacularly out of the ordinary, SV fluctuates around 175,
i've only seen a couple of chokes (mostly around map start), which
should be pretty acceptable, I think?
Is it in any way possible to have the server log this information, with
timestamps? Possibly to correlate against a chatlog, and see if
anything out of the ordinary happens when players call "lag"?
I am not sure what game you are using since you didn't say. That can
make a difference on how to read the information.
You might want to tell us what CPU you are trying to use. It may be
just too old. It seems that people often try to use old CPUs,
thinking that the server won't be CPU intensive, when it actually is.
Current srcds games like TF2 and CSS are very CPU intensive,
especially TF2. You need a fairly modern CPU. I have an old AMD x2
4600+ that runs a 24-player TF2 server and that's about as old and
slow as you would want to go.
Note that the server is not optimized for multi-core CPUs. Once the
individual CPU core in use is maxed out, that's it, performance will
suffer.
The default server-side FPS is 300, and it can be a sign of trouble
when it drops down below 60 and you start getting choke. Occasional
intermittent drops are normal (spawn waves are particularly hard on
the CPU). Note that if your FPS is less than 300 when nobody is on
the server, use that as your baseline instead (local kernel timer
resolution maybe less than FPS).
Unfortunately, there isn't much else that you will find on the
server-side to help you get insight to your server performance, if
you've already done your due diligence.
Start up your client, connect to your server, and use net_graph to
watch live performance info;
http://developer.valvesoftware.com/wiki/TF2_Network_Graph
Watch the "sv", "var" and "choke" fields. Since you are new to this,
do yourself a favor and COMPLETELY ignore the upper-area of the
graph.
It indicates activity, NOT performance.
Note that the sample image that Valve provides is of a very poorly
performing server with a client with significant latency, so it's not
a good example of proper performance. Go play on about a dozen
different servers and watch how net_graph behaves.
I wrote a long "net_graph by example" post on a private forum some
time back. I should make it public some day.
If anyone disagrees with anything above, feel free to say so. I
don't know everything and there are many others who have been here
much longer than I have.
Peter Reinhold wrote:
CPU In Out Uptime Users FPS Players
74.27 90368.22 237868.38 2077 46 76.76 24
Specifically, the CPU and FPS.
I have seen CPU lying on 99.x for a couple of stats commands close
together, but, the CPU load on the box is nowhere near full load on
even
a single core, so, can anyone explain exactly what it means?
Also, the FPS, I think I read here on the list that it is normal
that
the FPS fluctuates, according to how much is happening on the
server,
so, any idea if I can use that metric for any kind of performance
bughunt?
_______________________________________________
To unsubscribe, edit your list preferences, or view the list archives, please
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux
